Prediction of the effect of demographic features on online market using with machine learning methods

It is seen that many campaigns that companies providing online grocery shopping services put into practice on their online sales platforms to increase product sales and acquire new customers have not been successful. The study aims to present an exemplary method that companies can use for an effective online market strategy by focusing on the main reason behind this existing problem. Hence, this study also has the potential to make a significant contribution to the ongoing studies in the field of technology. Various data were collected from 394 online market users between the ages of 20-86 who chose to shop online through a survey conducted for this study. Then, using the obtained data and machine learning algorithms, a model has been put forward that allows determining the shopping tendencies of the personal care category during online grocery shopping, depending on the demographic characteristics of the online market users. Then, to create ideal modeling on these data, Decision Trees, K-Nearest Neighbors (KNN), Gradient Augmented Trees (GBT), Random Forest, and Logistic Regression algorithms, which are classification algorithms that provide the most efficient results, were applied. Finally, a comparison was made over AUC (Area under the Curve), recall, f1-score (f1-score), and precision results. Thus, as an outcome of the study, it was determined that the Logistic Regression algorithm gave the best performance with a 0.83 accuracy rate and 0.92 AUC value among the compared algorithms. According to the model result users who are under the age of 43-47, and intensively use the internet daily, and do not prefer to pay with a credit card at the door make more purchases from the personal care category while they are shopping online.. In addition, it has been determined that users who think that their private life is better protected and that they can buy products from each category shop more from this category. However, it has emerged that users who think they can touch or feel the products do not prefer to shop online in this category. When the results obtained are examined, it can be said that if the surveys containing questions about the demographic information and shopping preferences of the users are carried out periodically by the companies, this will enable them to determine their target audiences better. In this way, it is foreseen that unnecessary investments made in the field can be prevented.